We consider the problem of dealing with irrelevant votes when a multi-case classifier is built from an ensemble of binary classifiers. We show how run-off elections can be used to...
Human interaction proofs (HIPs) have become common place on the internet due to their effectiveness in deterring automated abuse of online services intended for humans. However, th...
Kumar Chellapilla, Kevin Larson, Patrice Y. Simard...
Abstract. This paper presents an architecture that enables the recognizer to learn incrementally and, thereby adapt to document image collections for performance improvement. We ar...
In typical classification tasks, we seek a function which assigns a label to a single object. Kernel-based approaches, such as support vector machines (SVMs), which maximize the ...
CAPTCHAs protect online resources and services from automated access. From an attacker’s point of view, they are typically perceived as an annoyance that prevents the mass creat...
Manuel Egele, Leyla Bilge, Engin Kirda, Christophe...